• Efficient and Deployable Click Fraud Detection for Mobile Applications

    Author(s):
    Prabhakara Uyyala (see profile)
    Date:
    2021
    Item Type:
    Article
    Permanent URL:
    https://doi.org/10.17613/8m7q-cm76
    Abstract:
    However, because the detection can be easily avoided, such approaches may suffer from a large number of false negatives. For example, when the clicks are behind proxies or globally scattered, such approaches may suffer from a high number of false negatives. In this work, we describe AdSherlock, a client-side (within the app) click fraud detection solution for mobile apps that is efficient and deployable. The computation-intensive activities of click request identification are split into an offline and an online method by AdSherlock. AdSherlock generates both accurate and probabilistic patterns based on URL (Uniform Resource Locator) tokenization in the offline approach. These patterns are employed in the online procedure for click request identification and, in combination with an ad request tree model, for click fraud detection. We build an AdSherlock prototype and test its performance with real-world apps. Through binary instrumentation, the online detector is inserted into the app executable archive. AdSherlock achieves higher click fraud detection accuracy than the state of the art, with low runtime overhead, according to the results.
    Metadata:
    Published as:
    Journal article    
    Status:
    Published
    Last Updated:
    10 months ago
    License:
    Attribution-NonCommercial

    Downloads

    Item Name: pdf 225-ijaema-january-2021.pdf
      Download View in browser
    Activity: Downloads: 1463